def test_keep_or_kill(self):
for _ in range(10):
freq = random.uniform(0.01, 0.4)
sr = 1024
tone = pure_tone(freq * sr, sr=sr, dur=10)
# For this test we accept 5% tolerance in amplitude, or -26dB in power.
tol = 5
zeros = 16
# If cutoff frequency is under freq, output should be input
y_pass = filters.highpass_filter(tone, 0.9 * freq, zeros=zeros)
self.assertSimilar(y_pass,
tone,
tone,
f"freq={freq}, pass",
tol=tol)
# If cutoff frequency is over freq, output should be zero
y_killed = filters.highpass_filter(tone, 1.1 * freq, zeros=zeros)
self.assertSimilar(y_killed,
0 * tone,
tone,
f"freq={freq}, kill",
tol=tol)
def test_fft_nofft(self):
for _ in range(10):
x = th.randn(1024)
freq = random.uniform(0.01, 0.5)
y_fft = filters.highpass_filter(x, freq, fft=True)
y_ref = filters.highpass_filter(x, freq, fft=False)
self.assertSimilar(y_fft, y_ref, x, f"freq={freq}", tol=0.01)
def forward(self, audio: torch.Tensor) -> torch.Tensor:
frequency = self.sample_uniform_frequency()
cutoff = self.cutoff_frequency(frequency)
audio = highpass_filter(audio, cutoff=cutoff)
return audio
def test_constant(self):
x = th.ones(2048)
for zeros in [4, 10]:
for freq in [0.01, 0.1]:
y_high = filters.highpass_filter(x, freq, zeros=zeros)
self.assertLessEqual(y_high.abs().mean(), 1e-6, (zeros, freq))
def test_same_as_highpass(self):
x = th.randn(1024)
y_ref = filters.highpass_filter(x, 0.2)
y = filters.bandpass_filter(x, 0.2, 0.5)
self.assertSimilar(y, y_ref, x)